Update cpupower's P-State frequency calculation and reporting with AMD
Family 1Ah+ processors, when using the acpi-cpufreq driver. This is due
to a change in the PStateDef MSR layout in AMD Family 1Ah+.
Tested on 4th and 5th Gen AMD EPYC system
Signed-off-by: Ananth Narayan <Ananth.Narayan@amd.com>
Signed-off-by: Dhananjay Ugwekar <Dhananjay.Ugwekar@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Pull turbostat updates from Len Brown:
- Survive sparse die id's seen in Linux-6.9
- Handle clustered-uncore topology in new/upcoming hardware
- For non-root use, add ability to see software C-state counters
- Enable reading core and package hardware cstate via perf, and prefer
perf over the MSR driver access for these counters
* tag 'turbostat-for-Linux-6.10-merge-window' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux:
tools/power turbostat: version 2024.05.10
tools/power turbostat: Ignore pkg_cstate_limit when it is not available
tools/power turbostat: Fix order of strings in pkg_cstate_limit_strings
tools/power turbostat: Read Package-cstates via perf
tools/power turbostat: Read Core-cstates via perf
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs
tools/power turbostat: Add columns for clustered uncore frequency
tools/power turbostat: Enable non-privileged users to read sysfs counters
tools/power turbostat: Replace _Static_assert with BUILD_BUG_ON
tools/power turbostat: Add ARL-H support
tools/power turbostat: Enhance ARL/LNL support
tools/power turbostat: Survive sparse die_id
tools/power turbostat: Remember global max_die_id
tools/power turbostat: Harden probe_intel_uncore_frequency()
tools/power turbostat: Add "snapshot:" Makefile target
Pull x86 platform driver updates from Hans de Goede:
- New drivers/platform/arm64 directory for arm64 embedded-controller
drivers
- New drivers:
- Acer Aspire 1 embedded controllers (for arm64 models)
- ACPI quickstart PNP0C32 buttons
- Dell All-In-One backlight support (dell-uart-backlight)
- Lenovo WMI camera buttons
- Lenovo Yoga Tablet 2 Pro 1380F/L fast charging
- MeeGoPad ANX7428 Type-C Cross Switch (power sequencing only)
- MSI WMI sensors (fan speed sensors only for now)
- Asus WMI:
- 2024 ROG Mini-LED support
- MCU powersave support
- Vivobook GPU MUX support
- Misc. other improvements
- Ideapad laptop:
- Export FnLock LED as LED class device
- Switch platform profiles using thermal management key
- Intel drivers:
- IFS: various improvements
- PMC: Lunar Lake support
- SDSI: various improvements
- TPMI/ISST: various improvements
- tools: intel-speed-select: various improvements
- MS Surface drivers:
- Fan profile switching support
- Surface Pro thermal sensors support
- ThinkPad ACPI:
- Reworked hotkey support to use sparse keymaps
- Add support for new trackpoint-doubletap, Fn+N and Fn+G hotkeys
- WMI core:
- New WMI driver development guide
- x86 Android tablets:
- Lenovo Yoga Tablet 2 Pro 1380F/L support
- Xiaomi MiPad 2 status LED and bezel touch buttons backlight
support
- Miscellaneous cleanups / fixes / improvements
* tag 'platform-drivers-x86-v6.10-1' of git://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86: (128 commits)
platform/x86: Add new MeeGoPad ANX7428 Type-C Cross Switch driver
devm-helpers: Fix a misspelled cancellation in the comments
tools arch x86: Add dell-uart-backlight-emulator
platform/x86: Add new Dell UART backlight driver
platform/x86: x86-android-tablets: Create LED device for Xiaomi Pad 2 bottom bezel touch buttons
platform/x86: x86-android-tablets: Xiaomi pad2 RGB LED fwnode updates
platform/x86: x86-android-tablets: Pass struct device to init()
platform/x86/amd: pmc: Add new ACPI ID AMDI000B
platform/x86/amd: pmf: Add new ACPI ID AMDI0105
platform/x86: p2sb: Don't init until unassigned resources have been assigned
platform/surface: aggregator: Log critical errors during SAM probing
platform/x86: ISST: Support SST-BF and SST-TF per level
platform/x86/fujitsu-laptop: Replace sprintf() with sysfs_emit()
tools/power/x86/intel-speed-select: v1.19 release
tools/power/x86/intel-speed-select: Display CPU as None for -1
tools/power/x86/intel-speed-select: SST BF/TF support per level
tools/power/x86/intel-speed-select: Increase number of CPUs displayed
tools/power/x86/intel-speed-select: Present all TRL levels for turbo-freq
tools/power/x86/intel-speed-select: Fix display for unsupported levels
tools/power/x86/intel-speed-select: Support multiple dies
...
New since 2024.04.08:
Len Brown (6):
tools/power turbostat: Add "snapshot:" Makefile target
tools/power turbostat: Harden probe_intel_uncore_frequency()
tools/power turbostat: Remember global max_die_id
tools/power turbostat: Survive sparse die_id
tools/power turbostat: Add columns for clustered uncore frequency
tools/power turbostat: version 2024.05.10
Patryk Wlazlyn (7):
tools/power turbostat: Replace _Static_assert with BUILD_BUG_ON
tools/power turbostat: Enable non-privileged users to read sysfs counters
tools/power turbostat: Avoid possible memory corruption due to sparse topology IDs
tools/power turbostat: Read Core-cstates via perf
tools/power turbostat: Read Package-cstates via perf
tools/power turbostat: Fix order of strings in pkg_cstate_limit_strings
tools/power turbostat: Ignore pkg_cstate_limit when it is not available
Zhang Rui (2):
tools/power turbostat: Enhance ARL/LNL support
tools/power turbostat: Add ARL-H support
Signed-off-by: Len Brown <len.brown@intel.com>
When running in no-msr mode, the pkg_cstate_limit is not populated, thus
we use perf to determine if given pcstate counter is present on the
platform.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Reading the counters via perf can be done in bulk with a single syscall,
making the counter values more accurate with respect to one another by
minimizing the time gap between individual counter reads.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Reading the counters via perf can be done in bulk with a single syscall,
making the counter values more accurate with respect to one another by
minimizing the time gap between individual counter reads.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Save the highest core and package id when parsing topology to
allocate enough memory when get_rapl_counters() is called with a core or
a package id as a domain.
Note that RAPL domains are per-package on Intel, but per-core on AMD.
Thus, the RAPL code effectively runs in different modes on those two
product lines.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
New machines have multiple uncore frequencies per package,
visible in /sys/devices/system/cpu/intel_uncore_frequency/uncore##/
turbostat now samples these frequencies each measurement interval.
For each package, turbostat now prints "UMHzX.Y" columns,
where X = domain_id, and Y = fabric_cluster_id.
The system summary for each UMHzX.Y column is the average value
for across all of the packages in the system.
Signed-off-by: Len Brown <len.brown@intel.com>
A group of counters called "sysfs" displays software
C-state request counts and resulting perceived C-state residency.
They are not built-in counters that turbostat knows about ahead of time,
rather they are discovered in sysfs when turbostat starts.
Thus, they are added dynamically, using the same interface
as user-added MSR counters.
When turbostat enters "no-msr" mode, such as when running as a
non-privileged user, it clears all added counters.
Updating that to clear only actual MSR added counters
allows regular users to see the sysfs counters.
[lenb: commit message]
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Add turbostat support for ARL-H, which behaves the same as ARL.
[lenb: also add ARL-U]
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
ARL/LNL don't have PC8, other than that, it behaves the same as CNL.
Copy cnl_features for ARL/LNL, except that PC8 support is removed.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
This version addresses issues with:
- Support of SST BF/TF support per level
- Increase number of CPUs displayed
- Present all TRL levels for turbo-freq
- Fix display for unsupported levels
- Support multiple dies
- Increase die count
- Change CPU display for non compute domain
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
SST BF and TF can be enabled/disabled per level. So check the current
level support from the mask of supported levels.
This change from a single level to mask for info.sst_tf_support and
info.sst_tf_support is indicated by API version change. Use as mask for
API version above 2. In this way there is no change in behavior when
running on older kernel with API version 2.
Since the tool can support now API version 3, update the supported API
version.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Currently max 128 CPUs can be displayed in the enable CPU list. Double
the range. Since the size is big for stack allocation, change to static.
Here changing to static is fine as these functions are called in serial.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
For turbo-freq feature, only 3 levels of frequencies are displayed even
if platform support more. Present all levels based on the CPU model.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
During call to "intel-speed-select turbo-freq info" some junk values are
reported for unsupported levels. Initialize the structure fact_info with
0s, so that isst_fact_display_information() will skip "0" values in the
frequency.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
When the die id is same as punit compute die ID, treat them same. In this
case, when for_each_online_power_domain_in_set() is called, then don't
loop for each punit in a die. Just loop for all punits in a package.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
There is request from the end user to print this field to better
query what type of update capability is supported on this platform.
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Turbostat assumed that every package had a die_id = 0.
When this assumption was violated, it exited
when looking for the package uncore frequency:
turbostat: /sys/.../intel_uncore_frequency/package_01_die_00/current_freq_khz: open failed: No such file or directory
Signed-off-by: Len Brown <len.brown@intel.com>
If sysfs directory "intel_uncore_frequency/cluster00/" exists,
then use uncore cluster code (now its own routine).
The previous check for
"intel_uncore_frequency/package_00_die_00/current_freq_khz",
could be unreliable in the face of sparse die id's.
Signed-off-by: Len Brown <len.brown@intel.com>
Kernel developers often need to diagnose remote customer systems
with the latest turbostat, yet customers are running binary distros
with out-dated turbostat and the customer has no experience
cloning linux kernel trees.
Add a turbostat "snapshot" makefile target to create a standalone
source snapshot from the developer's git tree, appropriately hacked
so that the customer can build turbostat without a kernel tree.
Include the turbostat binary in the snapshot, for convenience in
those situations where the source and destination are trusted,
(and have new enough glibc to execute).
The snapshot is named with the date it was taken rather than
the turbostat VERSION, as it could occur between VERSIONS...
Signed-off-by: Len Brown <len.brown@intel.com>
Pull turbostat updates from Len Brown:
- Use of the CPU MSR driver is now optional
- Perf is now preferred for many counters
- Non-root users can now execute turbostat, though with limited
functionality
- Add counters for some new GFX hardware
- Minor fixes
* tag 'turbostat-2024.04.10' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux: (26 commits)
tools/power turbostat: v2024.04.10
tools/power/turbostat: Add support for Xe sysfs knobs
tools/power/turbostat: Add support for new i915 sysfs knobs
tools/power/turbostat: Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz
tools/power/turbostat: Fix uncore frequency file string
tools/power/turbostat: Unify graphics sysfs snapshots
tools/power/turbostat: Cache graphics sysfs path
tools/power/turbostat: Enable MSR_CORE_C1_RES support for ICX
tools/power turbostat: Add selftests
tools/power turbostat: read RAPL counters via perf
tools/power turbostat: Add proper re-initialization for perf file descriptors
tools/power turbostat: Clear added counters when in no-msr mode
tools/power turbostat: add early exits for permission checks
tools/power turbostat: detect and disable unavailable BICs at runtime
tools/power turbostat: Add reading aperf and mperf via perf API
tools/power turbostat: Add --no-perf option
tools/power turbostat: Add --no-msr option
tools/power turbostat: enhance -D (debug counter dump) output
tools/power turbostat: Fix warning upon failed /dev/cpu_dma_latency read
tools/power turbostat: Read base_hz and bclk from CPUID.16H if available
...
Much of turbostat can now run with perf, rather than using the MSR driver
Some of turbostat can now run as a regular non-root user.
Add some new output columns for some new GFX hardware.
[This patch updates the version, but otherwise changes no function;
it touches up some checkpatch issues from previous patches]
Signed-off-by: Len Brown <len.brown@intel.com>
Xe graphics driver uses different graphics sysfs knobs including
/sys/class/drm/card0/device/tile0/gt0/gtidle/idle_residency_ms
/sys/class/drm/card0/device/tile0/gt0/freq0/cur_freq
/sys/class/drm/card0/device/tile0/gt0/freq0/act_freq
/sys/class/drm/card0/device/tile0/gt1/gtidle/idle_residency_ms
/sys/class/drm/card0/device/tile0/gt1/freq0/cur_freq
/sys/class/drm/card0/device/tile0/gt1/freq0/act_freq
Plus that,
/sys/class/drm/card0/device/tile0/gt<n>/gtidle/name
returns either gt<n>-rc or gt<n>-mc. rc is for GFX and mc is SA Media.
Enhance turbostat to prefer the Xe sysfs knobs when they are available.
Export gt<n>-rc via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz.
Export gt<n>-mc via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
On Meteorlake platform, i915 driver supports the traditional graphics
sysfs knobs including
/sys/class/drm/card0/power/rc6_residency_ms
/sys/class/drm/card0/gt_cur_freq_mhz
/sys/class/drm/card0/gt_act_freq_mhz
At the same time, it also supports
/sys/class/drm/card0/gt/gt0/rc6_residency_ms
/sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz
/sys/class/drm/card0/gt/gt0/rps_act_freq_mhz
/sys/class/drm/card0/gt/gt1/rc6_residency_ms
/sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz
/sys/class/drm/card0/gt/gt1/rps_act_freq_mhz
gt0 is for GFX and gt1 is for SA Media.
Enhance turbostat to prefer the i915 new sysfs knobs.
Export gt0 via BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz.
Export gt1 via BIC_SMA_mc6/BIC_SMAMHz/BIC_SMAACTMHz.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Graphics driver (i915/Xe) on mordern platforms splits GFX and SA Media
information via different sysfs knobs.
Existing BIC_GFX_rc6/BIC_GFXMHz/BIC_GFXACTMHz columns can be reused for
GFX.
Introduce BIC_SAM_mc6/BIC_SAMMHz/BIC_SAMACTMHz columns for SA Media.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Running turbostat on a 16 socket HPE Scale-up Compute 3200 (SapphireRapids) fails with:
turbostat: /sys/devices/system/cpu/intel_uncore_frequency/package_010_die_00/current_freq_khz: open failed: No such file or directory
We observe the sysfs uncore frequency directories named:
...
package_09_die_00/
package_10_die_00/
package_11_die_00/
...
package_15_die_00/
The culprit is an incorrect sprintf format string "package_0%d_die_0%d" used
with each instance of reading uncore frequency files. uncore-frequency-common.c
creates the sysfs directory with the format "package_%02d_die_%02d". Once the
package value reaches double digits, the formats diverge.
Change each instance of "package_0%d_die_0%d" to "package_%02d_die_%02d".
[lenb: deleted the probe part of this patch, as it was already fixed]
Signed-off-by: Justin Ernst <justin.ernst@hpe.com>
Reviewed-by: Thomas Renninger <trenn@suse.de>
Signed-off-by: Len Brown <len.brown@intel.com>
Graphics sysfs snapshots share similar logic.
Combine them into one function to avoid code duplication.
No functional change.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Graphics drivers (i915/Xe) have different sysfs knobs on different
platforms, and it is possible that different sysfs knobs fit into the
same turbostat columns.
Instead of specifying different sysfs knobs every time, detect them
once and cache the path for future use.
No functional change.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Some of the future Intel platforms will require reading the RAPL
counters via perf and not MSR. On current platforms we can still read
them using both ways.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
If user request --no-msr or is not able to access the MSRs,
turbostat should clear all the counters added with --add.
Because MSR access permission checks are done after the cmdline is
parsed, the decision has to be defered up until the transition into
no-msr mode happen.
Signed-off-by: Len Brown <len.brown@intel.com>
Checking early if the permissions are even needed gets rid of the
warnings about some of them missing. Earlier we issued a warning in case
of missing MSR and/or perf permissions, even when user never asked for
counters that require those.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
By using the perf API we spend less time in between the reads of the
counters, resulting in more accurate calculations of the dependent
metrics.
Using perf API is also usually faster overall, although cache miss, if
we get one, is more costly when using perf vs MSR driver.
We would fallback to the msr reads if the sysfs isn't there or when in
--no-perf mode.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Eliminate redundant debug output for core and package scope counters.
Include name and path for all "ADDED" counters.
Signed-off-by: Len Brown <len.brown@intel.com>
Previously a failed read of /dev/cpu_dma_latency erroneously complained
turbostat: capget(CAP_SYS_ADMIN) failed, try "# setcap cap_sys_admin=ep ./turbostat
This went unnoticed because this file is typically visible to root,
and turbostat was typically run as root.
Going forward, when a non-root user can run turbostat...
Complain about failed read access to this file only if --debug is used.
Signed-off-by: Len Brown <len.brown@intel.com>
Merge a cpupower utility documentation update for 6.9-rc1 from Shuah Khan:
"This cpupower update for Linux 6.9-rc1 consists of a single fix
to a typo in cpupower-frequency-info.1 man page."
* tag 'linux-cpupower-6.9-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux:
Fix cpupower-frequency-info.1 man page typo
cpupower-frequency-info.1 man page type is incorrect for
related-cpus. Fix it.
utils/cpufreq-info.c
{"related-cpus", no_argument, NULL, 'r'},
{"affected-cpus", no_argument, NULL, 'a'},
Fixed changelog before applying:
Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: Jan Kratochvil <jan@jankratochvil.net>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
In function get_pkg_num() if fopen_or_die() succeeds it returns a file
pointer to be used. But fclose() is never called before returning from
the function.
Signed-off-by: Samasth Norway Ananda <samasth.norway.ananda@oracle.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
